UKOLN is supported by:

1. Signed metadata : method and application International Conference on Dublin Core and Metadata Applications, 3 – 6 October 2006, Mexico Julie Allinson (presenter) Repositories Research Officer UKOLN, University of Bath Emma Tonkin (author) Interoperability Focus Officer UKOLN, University of Bath UKOLN is supported by: www.bath.ac.uk

2. Contents • Overview • Introduction • Brief background on digital signatures • Signing DC metadata – approaches and issues • Use cases

3. Overview • Why do we need to digitally sign metadata records? Currently, we (probably) don’t, but … increasing numbers of metadata providers + additional ways of reusing data = increasing issues of trust, provenance and identity • Digitally signing metadata records through a Public-Key Infrastructure (PKI) is one potential solution

4. Introduction • The current digital library world works on • Implicit trust - metadata providers are trusted because we ‘know’ them • Explicit trust – e.g. the OAI-PMH <provenance> tag provides information • A small network of trusted and responsible organisations • In the future, we can envisage • Metadata-enabled filesystems • Increased informal metadata tagging services • Larger-scale networks • More opportunities for abuse, e.g. spamming • Less accountability and responsibility (= less trust)

5. Digital signatures • Date back to 1976 • Use cryptographic techniques • Similar to handwritten signatures • Permit the verification of messages The most common solution is Public-Key Infrastructure (PKI)

6. PKI – how does it work? • A digital signer has 2 keys • Private key – used to create the signature • Public key – used by third-parties to verify the author • Public keys are distributed by a distribution system, e.g. a key server containing keys and identity information • PKI is useful in establishing a network of trust • … but it has limitations • It is possible to produce a key with fictitious, false or stolen identity

7. Signing Dublin Core metadata • Dublin Core is unusual in that it can be represented in different ways, e.g. XML, RDF, XHTML • Approaches • The XML Signatures standard provides flexible methods for signing and verifying data objects in XML • An XML metadata record could be wrapped within an XML Signature • OpenPGP is an alternative mechanism • OpenPGP could be used to sign the name-value pairs within a metadata record • A standardised approach is required for both mechanisms

8. XML Signatures (1) <Signature ID?>      <SignedInfo>        <CanonicalizationMethod/>        <SignatureMethod/>        (<Reference URI? >          (<Transforms>)?          <DigestMethod>          <DigestValue>        </Reference>)+      </SignedInfo>      <SignatureValue>     (<KeyInfo>)?     (<Object ID?>)* </Signature> • “XML digital signatures are represented by the Signature element which has the following structure (where "?" denotes zero or one occurrence; "+" denotes one or more occurrences; and "*" denotes zero or more occurrences)” (from http://www.w3.org/TR/xmldsig-core/ )

9. XML Signatures (2) • Reference URI='http://example.com/the-signed-dc-record'> <DigestMethod/> (the type of signature used) <DigestValue></DigestValue>   (contains the signature, an encrypted value)

10. <metadata   xmlns="http://example.org/myapp/"  xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"  xsi:schemaLocation="http://example.org/myapp/http://example.org/myapp/schema.xsd"   xmlns:dc="http://purl.org/dc/elements/1.1/">  <dc:title>     UKOLN   </dc:title>   <dc:description>     ...   </dc:description>   <dc:publisher>     UKOLN, University of Bath   </dc:publisher>   <dc:identifier> http://www.ukoln.ac.uk/  </dc:identifier> </metadata> <dsig:Signature       xmlns:dsig="http://www.w3.org/2000/09/xmldsig#"      xmlns:dsig-xpath="http://www.w3.org/2002/06/xmldsig-filter2">        <dsig:SignedInfo>          ...        <dsig:Reference URI="#metadata"> </dsig:Signature> (see http://dublincore.org/documents/dc-xml-guidelines/ )

11. Minimum components of a signature • XML Signatures is not the only method for signing metadata. The following information would be needed for any signature: • character set, encoding • current character encoding (to enable conversion) • signature method (e.g. sha1) • ID - the ID of the signer, could be an email address • the signature itself, analogously referred to in XML-Signature as 'digestvalue‘ • Information about the signed item, such as the metadata schema • for example, use of 'oai-dc‘ might be taken to mean 'expect all of the oai-dc elements to be present in key-value pairs, check the signature over all of them'

12. Issues in signing metadata • Changes in encoding and/or character set will invalidate signatures • A signature is also invalidated if changes are made to the metadata record. The ‘new’ package should be re-signed by whoever makes those changes and the original signed package is effectively lost. • How do we maintain the integrity of the original signature and the original metadata record? • Alternative methods • digital amendments or annotations appended outside of the original package • Information supplied by an OAI harvester might be signed by the OAI harvester (amendments/annotations) and/or by the repository (unchanged metadata). • Undersigning all metadata by the harvester offers a kind of ‘traceroute’ to show the history of that record • But it could lead to large packets of metadata being transferred around networks

13. Provenance in aggregation • Currently • aggregators are most likely to harvest content from the originating repository • In the future the repository ecology looks much more complex • increased repository numbers and sharing metadata between repositories • more aggregators and aggregation of content from more sources • increased availability of informal metadata sources • Current trust mechanisms (perceived integrity of the source) do not scale • PKI could be used to identify the origination of the metadata and its route through other repositories and/or aggregators

14. Potential applications • A public-key infrastructure adds complexity, resource and infrastructure overheads • It is valuable only where the functionality is explicitly required or provides clear advantages • Some examples • Provenance in aggregation • A distributed metadata cloud • Metadata handling and trust in mobile devices and ad-hoc networks

15. Distributed metadata cloud • = a loosely coupled, interoperating collection of heterogeneous metadata sources and other services • Information is seamlessly passed between members of the ‘cloud’ • Identifying provenance and identity provides • A trust mechanism for assessing the potential value of information • A verifiable transmission path and origin of annotations • Access to additional information about the data source

16. Mobile devices and ad hoc networks • In a centralised system it is relatively easy to ascertain the originator of information • … but with increasingly pervasive ad hoc Internet access • offered in a decentralised way • lightweight PKI can help identify each stage in the chain and thereby help us distinguish the spam from the trusted

17. Conclusion • Issues of provenance and identity are dealt with in the current digital library realm by the perceived integrity of a source • As the number of metadata sources and aggregators increase, these informal mechanisms may prove insufficient and metadata may be subject to abuse • Digitally signing metadata records can help to identify provenance • Public key infrastructure functionality offers particular cryptographic methods to digitally signing metadata • And can help to create new networks of trust